fix(vm_collect,vm_hot_plug): fix label selector and memory bugs

[stevefulme1]

Summary

• vm_collect role: Fix two copy-paste bugs from vm_backup_restore that break VM collection when using label_selectors
• vm_hot_plug role: Fix wrong condition in compute patch template that prevents memory-only hotplug

Bug Details

Bug 1 — Wrong variable names in task name (vm_collect/tasks/main.yml:48)

The label selector block's task name referenced vm_backup_restore_collect_obj and collect_obj_default_kind — leftover from the vm_backup_restore role. Both are undefined in the vm_collect context, causing a variable error when the block evaluates.

Fix: Changed to vm_collect_obj | default(vm_collect_obj_default_kind) to match the rest of the role.

Bug 2 — Wrong loop variable in set_fact (vm_collect/tasks/main.yml:64)

The set_fact in the label_selectors block referenced vm_collect_no_label_selectors_response (the loop_var from the other block) instead of vm_collect_label_selectors_response. This means VMs queried via label selectors were never added to the collection list — the variable was undefined, so the task would error or produce empty results.

Fix: Changed vm_collect_no_label_selectors_response → vm_collect_label_selectors_response.

Bug 3 — Wrong condition for memory hotplug (vm_hot_plug/templates/compute_patch.yml.j2:13)

The Jinja2 template checked 'cpu' in vm_hot_plug_compute before rendering the memory patch (/spec/template/spec/domain/memory/guest). This caused:

• Memory-only hotplug silently skipped (condition false)
• CPU-only hotplug to fail (tries to render vm_hot_plug_compute['memory'] which is undefined)

Fix: Changed condition to 'memory' in vm_hot_plug_compute.

Test plan

• Run vm_hot_plug with label_selectors in the request — VMs should now be collected and patched
• Run vm_hot_plug with CPU-only compute change — should patch CPU without attempting memory
• Run vm_hot_plug with memory-only compute change — should patch memory correctly
• Run vm_hot_plug with both CPU and memory — both patches applied
• Run vm_hot_plug with namespace/names (no label_selectors) — verify no regression

🤖 Generated with Claude Code

[stevefulme1]

Observed Scenario

A user reported CPU hotplug failing via AAP with the following job template vars:

vars:
  openshift_host: "{{ openshift_host }}"
  openshift_api_key: "{{ openshift_api_key }}"
  openshift_verify_ssl: "{{ openshift_verify_ssl | default(false) | bool }}"
  vm_hot_plug_request:
    - namespace: "{{ vm_namespace }}"
      names:
        - "{{ vm_name }}"
      compute:
        cpu:
          cores: "{{ cpu_cores | int }}"
      restartIfRequired: "{{ restart_if_needed | default(false) | bool }}"

Observed behavior:

• "Initialize Variables" runs fine (vm_hot_plug_vms: [])
• "Invoke Collect VM Role" skips — "skipped_reason": "No items in the list"
• "Process Hot Plug VM" skips — "skipped_reason": "No items in the list"

Analysis:

This scenario uses namespace + names (no label_selectors), so it takes the "without label selector" path in vm_collect. The three bugs fixed in this PR are real and would compound the problem:

1. Bug add ci workflow #1 (task name) — If Ansible strict mode or ANSIBLE_ACTION_WARNINGS is enabled, the undefined vm_backup_restore_collect_obj reference in the label selector block's task name could cause failures even when that block is skipped during template evaluation.
2. Bug fix: trigger first release with ci workflow #2 (wrong loop var) — Would break any request that does use label_selectors, since collected VMs would never be added to the list.
3. Bug chore: fix lint issues, add gitleaks pre-commit hook #3 (memory condition) — Would hit this user the moment they attempt memory-only hotplug or would cause a failure if they set CPU without memory (template tries to render undefined memory value).

For the specific "No items in the list" skip on vm_hot_plug_request, the root cause may also involve AAP variable precedence — the vars need to reach the vm_hot_plug role scope. If vm_hot_plug_request is defined at the playbook include_role vars level but the survey/extra vars substitution ({{ vm_namespace }}, {{ vm_name }}, etc.) isn't resolving, the list could evaluate to empty or the role could fall back to defaults/main.yml which defaults to [].

[sabre1041]

LGTM

[sabre1041]

LGTM

Omitted seeing conflicted file

@stevefulme1 would you be able to review the conflict in the test-requirements.txt file?

[stevefulme1]

Rebased onto main and resolved the test-requirements.txt conflict. The upstream already had the tox-ansible<26.2.2 pin so the conflict was just a redundant tox line from the earlier iterative fixes. Also resolved a tox-ansible.ini conflict where upstream had added sh and ade to allowlist_externals.

[sabre1041]

LGTM